,
8 tweets,
5 min read
Read on Twitter
In this interview by @lexfridman of @ylecun Lecun says the math of deep learning has more to do with Cybernetics than that found in computer science. Norbert Wiener formulated Cybernetics based on the idea of negative feedback.
@lexfridman @ylecun This was before electronic computers were invented. This was when Electrical Engineering was seeing a split into power engineering and control engineering. It's the latter that Wiener's work had its influence.
@lexfridman @ylecun Artificial Intelligence was coined 15 years later and Wiener was not invited. The A.I. proponents wanted to focus on the emerging tools of computer science and not the older tools of continuous mathematics.
@lexfridman @ylecun What was lost when GOFAI dominated was the Cybernetic realization that control in biology and machines share in common a negative feedback process. It is only in the Deep Learning revolution that revealed the significance of this profound insight from the 1940s
@lexfridman @ylecun When you look back into history, you can understand the appeal of the new paradigm of computer science and its potential to solve intelligence. This appeal is the same reason why GOFAI was so popular for half a century.
@lexfridman @ylecun However Wiener's thesis expressed in "Behavior, Purpose and Teleology" courses.media.mit.edu/2004spring/mas… captures the commonality of intelligence exhibited by both biological systems and human-engineered machines.
@lexfridman @ylecun What Wiener could not see was the rapid rise of digital computation. The cross-disciplinary flavor of Cybernetics was superseded by research in Complex Adaptive Systems. However, Cybernetics has always been associated with cognition and why it preserves its relevance today.
@lexfridman @ylecun @threadreaderapp unroll
